Towards Real Scene Super-Resolution with Raw Images

TL;DR

This paper introduces a novel pipeline for generating realistic training data and a dual CNN architecture to improve super-resolution performance on raw images, effectively recovering fine details in real-world scenarios.

Contribution

It proposes a new data generation pipeline and a dual CNN model that utilize raw image information for superior super-resolution in real scenes.

Findings

Raw data super-resolution recovers finer details.

The proposed pipeline produces more realistic training data.

The method outperforms existing approaches in real scenarios.

Abstract

Most existing super-resolution methods do not perform well in real scenarios due to lack of realistic training data and information loss of the model input. To solve the first problem, we propose a new pipeline to generate realistic training data by simulating the imaging process of digital cameras. And to remedy the information loss of the input, we develop a dual convolutional neural network to exploit the originally captured radiance information in raw images. In addition, we propose to learn a spatially-variant color transformation which helps more effective color corrections. Extensive experiments demonstrate that super-resolution with raw data helps recover fine details and clear structures, and more importantly, the proposed network and data generation pipeline achieve superior results for single image super-resolution in real scenarios.